Background Oral language skills are a critical foundation for literacy and more generally for educational success. The current study shows that oral language skills can be improved by providing suitable additional help to children with language difficulties in the early stages of formal education. Methods We conducted a randomized controlled trial with 394 children in England, comparing a 30‐week oral language intervention programme starting in nursery (N = 132) with a 20‐week version of the same programme starting in Reception (N = 133). The intervention groups were compared to an untreated waiting control group (N = 129). The programmes were delivered by trained teaching assistants (TAs) working in the children's schools/nurseries. All testers were blind to group allocation. Results Both the 20‐ and 30‐week programmes produced improvements on primary outcome measures of oral language skill compared to the untreated control group. Effect sizes were small to moderate (20‐week programme: d = .21; 30‐week programme: d = .30) immediately following the intervention and were maintained at follow‐up 6 months later. The difference in improvement between the 20‐week and 30‐week programmes was not statistically significant. Neither programme produced statistically significant improvements in children's early word reading or reading comprehension skills (secondary outcome measures). Conclusions This study provides further evidence that oral language interventions can be delivered successfully by trained TAs to children with oral language difficulties in nursery and Reception classes. The methods evaluated have potentially important policy implications for early education.